Could next gen consoles focus mainly on CPU?
- Thread starter Theeoo
- Start date
Well I believe Microsoft will do this with the xbox one, but Sony already stated they believe in classic Generations and that would mean PS5 Games are only for PS5. But maybe this time with building BC.
Ryzen is not that big anyway , one CCX is ~45mm2 irrc.
Jumping to another topic I wonder if the choice of the CPU for both the playstation and xbox had also to do with the impact on the process choice for the SOC (ad further down the line with shrinks).
Richland (bulldozer) low power version (35/45 Watts) were decent CPU definitely clockable within the PS/XB power budget and offering enough performance gains to justify a slight change in TDP or TDP balance (GPU/CPU). Now wasn't Jaguard done using library that made it "synthetizable" (??? my English is failing, I need to post more) design allow easier shrinks whereas Richland would have proven more bothering to port. Anyway old story, the CPU in this gen of console was pretty crappy but it was the best money could buy at the time.
I hope next gen we see proper CPU in the console, we have proper GPU, proper memory amount, it is the most serious bottleneck remaining on the glorified PC we call consoles.
Last edited:
phoenix_chipset
Regular
It certainly seems that way.
One thing to remember as well - jaguar wasn't really designed as an 8 core chip, so that puts some extra limits on it over the initial design in terms of thermals and clocks. I don't think Jag could hit 3hz in all honesty.
I also wonder if Nintendo hit a dead end in terms of Wii U's clocks, and how much further node shrinks could've carried it. I mean that chip was nearly microscopic at 45 nm haha.
HBRU
Regular
the main question is if a NEW gen will come in 2019 or (as I think) a PRO2... In this case maybe 16 jaguar cores will be there clocked as possible.... Something close to 10 billions transistor APU... 8 teraflops GPU... New gen I think will come in 2023 and will be quantum leap, not just an improvement....
look at the new HDR Ps4pro patch of the Witcher 3... A Pro2 is needed...
look at the new HDR Ps4pro patch of the Witcher 3... A Pro2 is needed...
Last edited by a moderator:
phoenix_chipset
Regular
As much as I'd love perfect hardware back compat, 16 jag cores would be pretty bad for a next gen... You can't just infinitely scale workload across cores at least not yet. Even if you could an 8 core Zen would still trash it.
I think if Sony were to release a pro 2 they would be pulling a Sega.
I think if Sony were to release a pro 2 they would be pulling a Sega.
D
Deleted member 13524
Guest
I thought the same about the XboneX.
Hey if they're launching a full year after the Pro then it's because they're definitely are waiting for Vega and Zen!
And then it came out with Jaguar cores and a Polaris GPU.
Why? You keep going with this argument but every console ever before hasn't had a mid-gen upgrade and has sold just fine and returned profits just fine even when the final games on the system struggled. PS2 owners didn't swap en masse to the XBox because it was much more powerful. PS2 didn't see all the previous PS1 owners change to PC in 1998/99 because it was old and slow technology. Unless you can present clear, hard evidence that a lack of a new console will see PS4 sales and profits plummet and Sony lose market share, you have zero case for a new console.
OneX was supposed to be a new generation
If there is a PRO2 it should be sold as a PS5
or new PS4... lol
Whatever it is called 5 or pro2, ultra, I guess we agree that Sony should launch it as soon as possible doable the goal is clearly to push out a cost efficient XB1X that can be shrunk further down the road. Quad core Zen, Navy GPU, 256 bit bus, 12-16GB of RAM and a nicer power profile than the XB1X.
TFLOPS figure should not be much of concern vs the efficiency of the GPU and its feature set, pushing the bar too high may turn into in an issue if Sony cleverly stick to its incremental upgrade strategy, too high a step could trigger more issues than it solves:
No room to play on pricing if needs be, tough to provide something significantly better in the upcoming years=> it may feel like securing a higher spot but it is also closing some option to evolve your product line and adapt your strategy.
We got to change the way we see thing as console are really close to close box PC and software is designed to be portable easily.
Beyond a point, pushing Jaguar clocks up will use more power per (unit of comparable work) than Zen. And given the huge strides made in power optimisation and power management over the last few years, I think Zen might already be there at the 2.1 and 2.3 gHz speeds seen in the Pro and X.
It's entirely possible that at the targeted performance levels of X2 and PS5, Zen would actually use less power and give you more power to spend on other areas of the system.
Zen is only going to get better with regards to IPC and perf/Watt ... and probably better in terms of perf/mm^2 too. Jaguar meanwhile is dead architecturally. With X1X Jaguar MS were able to tweak cache operation, latencies, reduce contention on memory accesses ... but the operation of the core itself stayed the same. There's probably not much more they can reasonably do to increase IPC, all they can do is push clock speeds higher on a smaller process, but that has its limits. Zen meanwhile will keep improving across the board.
It's entirely possible that at the targeted performance levels of X2 and PS5, Zen would actually use less power and give you more power to spend on other areas of the system.
Zen is only going to get better with regards to IPC and perf/Watt ... and probably better in terms of perf/mm^2 too. Jaguar meanwhile is dead architecturally. With X1X Jaguar MS were able to tweak cache operation, latencies, reduce contention on memory accesses ... but the operation of the core itself stayed the same. There's probably not much more they can reasonably do to increase IPC, all they can do is push clock speeds higher on a smaller process, but that has its limits. Zen meanwhile will keep improving across the board.
Zen transistor counts include various buses and interconnects, security stuff, memory controllers and probably all kinds of other stuff. Which you either won't need on console, or you'll have anyway. It's the transistor count of the CCX that matters. And the console version will likely use the half size L3 cache (4MB as opposed to 8). I don't know what the transistor count of the APU version of the CCX is, but going from die area, a 7nm version is not going to be prohibitive. The entire Ryzen 5 APU clocks in at less than 5 billion transistors.
A single APU type (reduced cache/area) CCX could far exceed the performance of 8 Jaguar cores.
Not without more substantial cooling and power delivery. Realistically the 3x PPC750 scheme was dead on arrival because of it's limited SIMD capability and only ~1.2 GHz clock speed. Nintendo should've just figured out a way to integrate the Broadway core either into the main system hub (which had pretty much everything aside from the Espresso CPU), or onto the CPU die, but still have more modern CPU cores to run actual Wii U games. There was a bewildering number of PPC cores out there to make use of.
Last edited:
Well, something was certainly weird when you've got an asymmetric L2 built with a more expensive eDRAM process as the presumably cheaper path for R&D. eDRAM may be denser than SRAM, but the chip still ended up with a ton of dead space due to pad limits anyway.
Are you referring to my post about the Wii U? I love how IBM said the Wii U had Watson/Power7 technology, which really was just their breakthrough in eDRAM density at the time.
With the XB1X tossing out the ESRAM, and Sony never using it in the first place this gen, specialized large cache systems are probably going to go the way of the dodo. Developers like the ease of big bandwidth + big memory, though I could see AMD gravitating towards larger L3 in their PC APUs to deal with bandwidth woes if it takes so long to bring 3D stacked memory.
